Study On Goat Motion Behavior Classification And Recognition Based On Improved Xgboost Algorithm

At present,goat breeding in Inner Mongolia has gradually changed from traditional stocking type to large-scale breeding mode.Among them,the monitoring of goat health and growth status still depends on artificial breeding experience,mainly using artificial observation of goat motion behavior for identification and judgment,which is not only high labor cost and high work intensity,but also difficult to meet the needs of large-scale breeding.Aiming at the problems of laborious observation,low efficiency and difficult to guarantee accuracy in manual monitoring,this paper designs an intelligent monitoring system for goat’s motion behavior based on three-axis acceleration sensor,which can monitor goat’s motion behavior in real time and automatically classify goat’s motion behavior.Through the identification of goat motion behavior,it can provide data reference for goat health assessment and intelligent disease early warning,and improve the economic value of goats.The specific work of this study is as follows: Firstly,to obtain the goat motion behavior data and provide data support for the subsequent goat motion behavior classification and prediction model,a set of goat motion behavior monitoring equipment based on three-axis acceleration sensor is designed.The device is fixed on the back of the goat,the motion behavior data of the goat is collected in real time,and the real-time collected data is transmitted to the cloud storage through the wireless transmission module.The four running behaviors of lying,standing,walking and running were marked by video calibration,and the original data were denoised by wavelet denoising method.Secondly,the classification and prediction model of goat sports behavior based on XGBoost is established.The parameters of XGBoost model are optimized by controlling variable method,and the parameters of XGBoost classification model are determined.Aiming at the problem that the classification prediction effect of XGBoost model is poor,the whale optimization algorithm(WOA)integrating social learning(SL)strategy is used to optimize the learning rate,maximum tree depth and leaf node weight of XGBoost model,and a classification prediction method of goat motion behavior based on SL-WOA-XGBoost is proposed.The experimental results show that the improved XGBoost can achieve a good classification of goat sports behavior.The single prediction accuracy of lying,standing,walking and running is 96.21%,86.69%,95.89% and 94.53% respectively.The average recognition rate of the four behaviors is 93.49%,which is 4.61% higher than that before the improvement.